Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.forth > #13928

Re: x64 assembler & RIP for call/jmp absolute indirect

Show all headers | View raw

"Alex McDonald" <blog@nospicedham.rivadpm.com> wrote in message
news:5aec6524-87a3-49ba-89bc-cca28a61d453@fr28g2000vbb.googlegroups.com...
> (posted to c.l.forth and c.l.asm.x86)
>
> While constructing a Forth assembler for x64 mode, I've added RIP
> addressing support, and come across a slight oddball encoding during
> testing the encodings. I've not yet been able to test this under an
> x64 system (that's what the assembler is for...) so I'm looking for
> anyone that can shed some light on the possible consequences of this.
> (Curly braces { } in the code below indicate a memory address; the
> equivalent of [$4111C] or [rip+$4111C] in other assemblers.)
>
> STC Experimental 32bit: 0.06.02 Build: 156
> mode64/32  ok ( set 64bit assembly )
> code x  ok
>   call { rip $4111C }   ok
>   call { $4111C }   ok
> next;  ok
> see x
> : x ( ? -- ? )
> \ std call compiles; code=$41B772 len=13 type=20
> \ defined in (console)
> ( $0 )    call    qword { rip $4111C }              \ FF154C114000
> ( $6 )    call    qword { $4111C }                  \ FF14254C114000
> ( $D )    ret                                       \ C3 ( end ) ok
>
> The first CALL is an "absolute indirect" to an 8byte address at
> absolute $4111C. The second is an explicit RIP encoding to an 8byte
> address at RIP+$4111C.
>
> Is this interpretation correct?
>

I left c.l.a.x. off.  Lately, you're lucky if you get one reply there.  So,
I won't consume your c.l.a.x. response, if it ever happens.

BTW, is your address encoding correct?

Your ORG appears to be zero, but if I disassemble the hex bytes you posted,
they seem to be at an address other than 0x4111C:

00000000  FF154C114000      call [rel 0x401152]
00000006  FF14254C114000    call [0x40114c]

I didn't check the manuals for the instruction encodings.  But, these appear
to be the address bytes (little-endian,i.e., reverse...) from each
respectively:

00 40 11 4C  (relative i.e., +6 bytes = 0x401152)
00 40 11 4C  (absolute 0x40114c)

So, it seems they're both encoded for address 0x40114C instead of 0x4111C...

(That's with NDISASM from NASM 2.10.01 - current version.)


Trying this NASM code:

BITS 64
ORG 0

CALL QWORD [ REL sym ]
CALL QWORD [ sym ]

times 266511 db 0
sym: ;0x4111C

Disassembling that gives:

00000000  FF1516110400      call [rel 0x4111c]
00000006  FF14251C110400    call [0x4111c]
...

That looks more like $4111C to me.

The address bytes (little-endian):

00 04 11 16 + 6 (relative to end of current instruction)
00 04 11 1C + 0 (absolute)


Rod Pemberton

Back to comp.lang.forth | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

x64 assembler & RIP for call/jmp absolute indirect Alex McDonald <blog@nospicedham.rivadpm.com> - 2012-07-11 09:25 -0700
  Re: x64 assembler & RIP for call/jmp absolute indirect Alex McDonald <blog@nospicedham.rivadpm.com> - 2012-07-11 10:32 -0700
  Re: x64 assembler & RIP for call/jmp absolute indirect mhx@iae.nl (Marcel Hendrix) - 2012-07-11 20:47 +0200
    Re: x64 assembler & RIP for call/jmp absolute indirect Alex McDonald <blog@rivadpm.com> - 2012-07-11 12:31 -0700
  Re: x64 assembler & RIP for call/jmp absolute indirect "Rod Pemberton" <do_not_have@notemailnot.cmm> - 2012-07-13 03:35 -0400
    Re: x64 assembler & RIP for call/jmp absolute indirect "Rod Pemberton" <do_not_have@notemailnot.cmm> - 2012-07-13 03:44 -0400
      Re: x64 assembler & RIP for call/jmp absolute indirect Alex McDonald <blog@rivadpm.com> - 2012-07-13 03:34 -0700

csiph-web